Françoise Leurs

A dynamic recurrent neural network for multiple muscles electromyographic mapping to elevation angles of the lower limb in human locomotion.

This paper describes the use of a dynamic recurrent neural network (DRNN) for simulating lower limb coordination in human locomotion. The method is based on mapping between the electromyographic signals (EMG) from six muscles and the elevation angles of the three main lower limb segments (thigh, shank and foot). The DRNN is a fully connected network of 35 hidden units taking into account the temporal relationships history between EMG and lower limb kinematics. Each EMG signal is sent to all 35 units, which converge to three outputs. Each output neurone provides the kinematics of one lower limb segment. The training is supervised, involving learning rule adaptations of synaptic weights and time constant of each unit. Kinematics of the locomotor movements were recorded and analysed using the opto-electronic ELITE system. Comparative analysis of the learning performance with different types of output (position, velocity and acceleration) showed that for common gait mapping velocity data should be used as output, as it is the best compromise between asymptotic error curve, rapid convergence and avoidance of bifurcation. Reproducibility of the identification process and biological plausibility were high, indicating that the DRNN may be used for understanding functional relationships between multiple EMG and locomotion. The DRNN might also be of benefit for prosthetic control.

Optimal walking speed following changes in limb geometry

SUMMARY The principle of dynamic similarity states that the optimal walking speeds of geometrically similar animals are independent of size when speed is normalized to the dimensionless Froude number (Fr). Furthermore, various studies have shown similar dimensionless optimal speed (Fr ∼0.25) for animals with quite different limb geometries. Here, we wondered whether the optimal walking speed of humans depends solely on total limb length or whether limb segment proportions play an essential role. If optimal walking speed solely depends on the limb length then, when subjects walk on stilts, they should consume less metabolic energy at a faster optimal speed than when they walk without stilts. To test this prediction, we compared kinematics, electromyographic activity and oxygen consumption in adults walking on a treadmill at different speeds with and without articulated stilts that artificially elongated the shank segment by 40 cm. Walking on stilts involved a non-linear reorganization of kinematic and electromyography patterns. In particular, we found a significant increase in the alternating activity of proximal flexors–extensors during the swing phase, despite significantly shorter normalized stride lengths. The minimal metabolic cost per unit distance walked with stilts occurred at roughly the same absolute speed, corresponding to a lower Fr number (Fr ∼0.17) than in normal walking (Fr ∼0.25). These findings are consistent with an important role of limb geometry optimization and kinematic coordination strategies in minimizing the energy expenditure of human walking.

Movement gating of beta/gamma oscillations involved in the N30 somatosensory evoked potential.

Evoked potential modulation allows the study of dynamic brain processing. The mechanism of movement gating of the frontal N30 component of somatosensory evoked potentials (SEP) produced by the stimulation of the median nerve at wrist remains to be elucidated. At rest, a power enhancement and a significant phase‐locking of the electroencephalographic (EEG) oscillation in the beta/gamma range (25–35 Hz) are related to the emergence of the N30. The latter was also perfectly identified in presence of pure phase‐locking situation. Here, we investigated the contribution of these rhythmic activities to the specific gating of the N30 component during movement. We demonstrated that concomitant execution of finger movement of the stimulated hand impinges such temporal concentration of the ongoing beta/gamma EEG oscillations and abolishes the N30 component throughout their large topographical extent on the scalp. This also proves that the phase‐locking phenomenon is one of the main actors for the N30 generation. These findings could be explained by the involvement of neuronal populations of the sensorimotor cortex and other related areas, which are unable to respond to the phasic sensory activation and to phase‐lock their firing discharges to the external sensory input during the movement. This new insight into the contribution of phase‐locked oscillation in the emergence of the N30 and in its gating behavior calls for a reappraisal of fundamental and clinical interpretation of the frontal N30 component. Hum Brain Mapp 2009.

Recognition of the physiological actions of the triphasic EMG pattern by a dynamic recurrent neural network

Triphasic electromyographic (EMG) patterns with a sequence of activity in agonist (AG1), antagonist (ANT) and again in agonist (AG2) muscles are characteristic of ballistic movements. They have been studied in terms of rectangular pulse-width or pulse-height modulation. In order to take into account the complexity of the EMG signal within the bursts, we used a dynamic recurrent neural network (DRNN) for the identification of this pattern in subjects performing fast elbow flexion movements. Biceps and triceps EMGs were fed to all 35 fully-connected hidden units of the DRNN for mapping onto elbow angular acceleration signals. DRNN training was supervised, involving learning rule adaptations of synaptic weights and time constants of each unit. We demonstrated that the DRNN is able to perfectly reproduce the acceleration profile of the ballistic movements. Then we tested the physiological plausibility of all the networks that reached an error level below 0.001 by selectively increasing the amplitude of each burst of the triphasic pattern and evaluating the effects on the simulated accelerating profile. Nineteen percent of these simulations reproduced the physiological action classically attributed to the 3 EMG bursts: AG1 increase showed an increase of the first accelerating pulse, ANT an increase of the braking pulse and AG2 an increase of the clamping pulse. These networks also recognized the physiological function of the time interval between AG1 and ANT, reproducing the linear relationship between time interval and movement amplitude. This task-dynamics recognition has implications for the development of DRNN as diagnostic tools and prosthetic controllers.

Physiological modules for generating discrete and rhythmic movements: action identification by a dynamic recurrent neural network

In this study we employed a dynamic recurrent neural network (DRNN) in a novel fashion to reveal characteristics of control modules underlying the generation of muscle activations when drawing figures with the outstretched arm. We asked healthy human subjects to perform four different figure-eight movements in each of two workspaces (frontal plane and sagittal plane). We then trained a DRNN to predict the movement of the wrist from information in the EMG signals from seven different muscles. We trained different instances of the same network on a single movement direction, on all four movement directions in a single movement plane, or on all eight possible movement patterns and looked at the ability of the DRNN to generalize and predict movements for trials that were not included in the training set. Within a single movement plane, a DRNN trained on one movement direction was not able to predict movements of the hand for trials in the other three directions, but a DRNN trained simultaneously on all four movement directions could generalize across movement directions within the same plane. Similarly, the DRNN was able to reproduce the kinematics of the hand for both movement planes, but only if it was trained on examples performed in each one. As we will discuss, these results indicate that there are important dynamical constraints on the mapping of EMG to hand movement that depend on both the time sequence of the movement and on the anatomical constraints of the musculoskeletal system. In a second step, we injected EMG signals constructed from different synergies derived by the PCA in order to identify the mechanical significance of each of these components. From these results, one can surmise that discrete-rhythmic movements may be constructed from three different fundamental modules, one regulating the co-activation of all muscles over the time span of the movement and two others elliciting patterns of reciprocal activation operating in orthogonal directions.

Rhythmic muscular activation pattern for fast figure-eight movement.

OBJECTIVE To address the question of how the CNS generates muscle activation patterns for complex gestures, we have chosen to study a figure-eight movement. We hypothesized that the well defined rhythmic aspect of this figure will provide further insights into the temporal features of multi-muscular commands. METHODS Subjects performed, as fast as possible, figure-eights initiated in the center of the figure with 4 different initial directions and 2 positions of the shoulder. We extracted the temporal modulation of the EMG patterns by calculating conjugate cross-correlation functions. RESULTS (1) The muscular command was tuned with respect to the rotational direction of the figure-eight, (2) two sets of synergistic muscles acted in a reciprocal mode, and (3) these reciprocal commands presented an invariant temporal correlation with the spatial component of the velocity having the highest frequency. CONCLUSION Our results suggest that the rhythmic features of certain drawing movements favor the partitioning of the muscles into synergistic groups acting in a reciprocal mode. The inclusion of an individual muscle in one group or the other takes into account the expected number of changes of direction in the movement as a whole. SIGNIFICANCE Muscular temporal synergies may depend on the rhythmic features of the trajectory.

Mu and alpha EEG rhythms during the arrest reaction in microgravity

Mu and alpha oscillations (8–12 Hz) are the most prominent electroencephalographic rhythms observed in awake, relaxed subjects. Different cortical sources may participate in these oscillations and appear to be modulated by the sensorimotor context and functional demands. In microgravity, the marked reduction in multimodal graviceptive inputs to cortical networks participating in the representation of space could be expected to affect these spontaneous rhythms. Here, we report the results of an experiment conducted over the course of 3 space flights, in which we quantified the power of the mu and alpha rhythms in relation to the arrest reaction (i.e. in 2 distinct physiological states: eyes open and eyes closed). We observed that the power of the spontaneous mu and alpha rhythms recorded in the eyesclosed state in the sensorimotor areas (mu rhythm) and in the parieto-occipital cortex (alpha rhythm) increased in microgravity. The suppression coefficient produced by eye-opening/ closure state transition also increased in microgravity. These results are discussed in terms of current theories on the source and the physiological significance of these EEG rhythms.

Development and Motor Control: From the First Step on

For performing their very first unsupported steps, often considered as a ‘milestone’ event in locomotor development, toddlers must find a compromise between at least two requirements: (1) the postural stability of the erect posture integrating the direction of gravity and (2) the dynamic control of the body and limbs for forward progression these two aspects. In adults, a series of experimental studies have provided evidence for coordinative laws that lead to a reduction of kinematic degrees of freedom. When the elevation angles of the thigh, shank and foot are plotted one versus the others, they describe a regular gait loop which lies to a plane. The plane orientation and the loop shape reflect the phase relationship between the different segments and therefore the timing of intersegmental coordination. The general pattern of intersegmental coordination and the stabilization of the trunk with respect of vertical are immature at the onset of unsupported walking in toddlers, but they develop in parallel very rapidly in the first few weeks of walking experience. Adult-like cross-correlation function parameters were reached earlier for shank-foot pairs than for thigh-shank indicating disto-proximal maturation of the lower limb segments coordination. We also demonstrated that a dynamic recurrent neural network (DRNN) is able to reproduce lower limb kinematics in toddler locomotion by using multiple raw EMG data. In the context of motor learning the DRNN may be considered as a model of biological learning mechanisms underlying motor adaptation. Using this artificial learning during the very first steps we found that the attractor states reached through learning correspond to biologically interpretable solutions.

A dynamic recurrent neural network for drawing multi-directional trajectories

This study describes the use of a dynamic recurrent neural network (DRNN) for simulating upper limb drawing movement in human. In a precedent work, Cheron et al. (1996) demonstrated that after learning, the DRNN has identified the preferential direction of the physiological action of the studied muscles. This suggests that the information of raw EMG signals is largely representative of the kinematics stored in the central motor pattern. The aim of this study was (1) to verify if the DRNN is able to recognize from EMG activities the kinematics of the same trajectory realized with different external coordinates and (2) if it is able to identify a difference in the EMG activities corresponding to the same trajectories but realized with a different intrinsic coordinates.

Reproducibility of the identification process of stump muscle EMG in prosthetic gait by a dynamic recurrent neural network

Dynamic recurrent neural networks (DRNN) have been succesfully used for identification of electromyography and arm trajectory relationship during complex movements, or for mapping between lower limb EMG and elevation angles of the thigh, shank and foot during healthy adults locomotion. The functional identification by the DRNN of muscle roles was clearly not random, but showed interesting interpretations of muscle activities on distant segments, considering both dynamical linkage of the segments and ground reaction forces during gait (Cheron et al. 2003). This study describes the use of a DRNN for simulating prosthetic lower limb coordination in transfemoral amputee’s gait from multiple muscles electromyographic signals (EMG) of the trunk, pelvis and stump. We tend to verify whether this simulator will also be able to identify the motion of a prosthetic limb driven by the remaining muscle groups.